| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2018-07-03 | |||
| 13:29:55 | dtantsur | "Dmitry used Microversions and you won't believe what happened next..." AAAaaaaAAAAaaa :D | |
| 13:30:45 | mordred | :) | |
| 13:31:25 | mordred | that one is growing on me | |
| 13:31:43 | dtantsur | mordred: I think #3 (unbearable lightness) is at the right crossing of being fun and acceptable for the audience :) | |
| 13:31:55 | dtantsur | but yes, #2 is so great | |
| 13:32:13 | mordred | \o/ | |
| 13:32:30 | mordred | also - white trash fast food is currently closed, so it's maybe not a great metaphor | |
| 13:33:22 | dtantsur | mordred: is it some known thing? | |
| 13:34:26 | dtantsur | I mean, this fast food thing | |
| 13:34:49 | mordred | dtantsur: http://www.whitetrashfastfood.com/ - it's my favorite berlin restaurant (or, it was) | |
| 13:35:06 | dtroyer | frickler: will do… I'm wrapping up osc-lib first then back to osc itself | |
| 13:35:38 | dtantsur | mordred: oh that's pity | |
| 13:36:10 | mordred | yah. I'm hoping it comes back | |
| 13:37:00 | dtantsur | yep. it's not even too far from my place (but in a, well, "funny" neighborhood) | |
| 13:37:14 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Only send force parameter to live migration if supported https://review.openstack.org/578960 | |
| 13:39:20 | frickler | dtroyer: cool, thx. I also asked a question about handling microversions, not sure whether this has been discussed globally already. mordred and dtantsur might also be interested. https://review.openstack.org/#/c/577768/4/openstackclient/compute/v2/keypair.py@84 | |
| 13:39:53 | dtantsur | s/this/his/ | |
| 13:40:09 | Luzi | mordred, I have been looking through openstacksdk, but i don't think the openstackcloud class is a suitable place for the signature generator. Where do you think we can add the signature generator class? | |
| 13:40:32 | dtantsur | frickler: well, I left the same comment on an earlier revision | |
| 13:41:58 | mordred | I think that's about as good as one can do before we plumb all the discovery stuff through - since the model there is assuming a configured api version | |
| 13:42:22 | frickler | dtantsur: oh, that was on the other review. but good to see we seem to agree | |
| 13:42:29 | dtantsur | oh, right, another one | |
| 13:42:44 | dtantsur | mordred: cannot we do version = max('2.2', os_compute_api_version) here? | |
| 13:43:02 | mordred | not without checking to see what api versions the remote side supports | |
| 13:43:05 | dtantsur | another thing to consider: we ended up with defaulting --os-baremetal-api-version to the known latest in ironicclient | |
| 13:43:30 | dtantsur | mordred: why not? the requested will fail reasonably if the remote side only supports 2.1, no? | |
| 13:44:28 | mordred | ah - hrm. yeah - I suppose so in this case | |
| 13:45:01 | dtantsur | assuming we can provide a microversion per request in OSC | |
| 13:45:20 | mordred | we can using sdk - not so much with novaclient | |
| 13:45:44 | dtantsur | which one is used by OSC? | |
| 13:45:47 | mordred | *except* - at least in my world, if a user has explicity set a config value for a specific microversion in compute_api_version - I would argue that we should not override that setting | |
| 13:46:23 | mordred | dtantsur: novaclient for now - although at some point we're hoping to get things shifted over to sdk - dean is waiting on a 1.0 from sdk before starting in on that | |
| 13:46:28 | dtantsur | yep, so maybe if os_compute_api_version is None: os_compute_api_version = '2.2' elif os_compute_api_version < '2.2': boom | |
| 13:46:51 | mordred | dtantsur: yah- I think that would be good and upwards compatible with a future where compute_api_version might not be set | |
| 13:47:17 | mordred | (today it'll always be set because of defaults - but once we've got the discovery stuff finished plumbed through, it can start being None and things working) | |
| 13:47:45 | dtantsur | ah, random defaults, similar to 1.6 in ironic? :) | |
| 13:47:48 | mordred | yup | |
| 13:48:01 | dtantsur | * 1.9 in ironicclient and 1.6 in openstacksdk because we LOVE consistency | |
| 13:48:08 | mordred | \o/ | |
| 13:48:37 | mordred | well - I've almost got the last discovery patch for sdk written in my head - so hopefully we're almost there | |
| 13:48:57 | mordred | Luzi: hello! yes - I agree, putting it in OpenStackCloud is not the right place ... | |
| 13:49:23 | mordred | Luzi: I've got a half-written patch on my laptop which should make a better place for you to put the generator | |
| 13:49:43 | mordred | Luzi: let me push up the work-in-progress in just a few minutes and see what you think | |
| 13:50:21 | Luzi | mordred, that sounds good :) | |
| 13:51:25 | mordred | Luzi: the generator class itself should probably just go in its own file in openstack/image - and then it should get attached to the image upload code in openstack/image/v1/_proxy and openstack/image/v2/_proxy ... but step 1 is to take the existing image upload logic from openstackcloud and move it to the proxy classes | |
| 13:58:45 | mordred | frickler: speaking of osc patches - https://review.openstack.org/#/c/540541/ | |
| 14:00:18 | dtantsur | mordred: any reason Resource does not implement the dict interface? This way it would be compatible with a Munch | |
| 14:00:33 | dtantsur | e.g. I could start returning openstack.baremetal.v1.Node from *machine* calls | |
| 14:00:56 | mordred | dtantsur: that is the _other_ main patch I wanna finish before cutting 1.0 | |
| 14:01:04 | dtantsur | very good :) | |
| 14:01:46 | mordred | dtantsur: https://review.openstack.org/#/c/530835/ is the very broken WIP patch for that | |
| 14:01:55 | dtantsur | I want to move *machine* calls to use openstack.baremetal.v1.* stuff internally, but currently they return incompatible result | |
| 14:02:07 | mordred | yup | |
| 14:02:13 | mordred | we share the same goal | |
| 14:03:18 | dtantsur | good | |
| 14:03:34 | dtantsur | I guess currently I can do munchify(node.to_dict()) as an ugly hack | |
| 14:04:00 | dtantsur | but what I'm hitting with my provisioning state patch is that I get a redundant GET call on each conversion between a Munch and a Node | |
| 14:04:07 | dtantsur | which is probably fine for now, but has to be fixed eventually | |
| 14:04:16 | mordred | yah - extra calls seem lame | |
| 14:04:26 | dtantsur | is it acceptable temporary? | |
| 14:06:13 | mordred | dtantsur: I'd be curious to discover why a GET call is being made during the conversion ... | |
| 14:07:12 | openstackgerrit | Monty Taylor proposed openstack/osc-lib master: Use openstack.config for config and Session objects https://review.openstack.org/579139 | |
| 14:07:13 | dtantsur | mordred: https://review.openstack.org/#/c/570895/4/openstack/cloud/openstackcloud.py@9400 causes uuid-to-Node conversion | |
| 14:07:28 | mordred | dtroyer: ^^ that patch passes for me locally now, so might be not a waste of your time to review | |
| 14:07:37 | dtantsur | even when this call is used from e.g. register_machine where we already have a Munch | |
| 14:07:55 | dtantsur | because we cannot pass in a Munch, we need a UUID or a Node | |
| 14:09:55 | mordred | if we have a munch, we should be able to make a Node from it without an api call - I'd think Node(**munch_node) would do the trick? | |
| 14:10:47 | dtantsur | good idea, maybe we should generalize it as Resource.from_munch? | |
| 14:11:03 | dtantsur | defaulting to just cls(**munch) | |
| 14:12:55 | frickler | mordred: oh, that makes me wonder whether it'll fix my neutron ssl issues. will check. | |
| 14:13:49 | mordred | dtantsur: good idea - although maybe make it Resource._from_munch for now - since we might not need it once we finish and land the other thing | |
| 14:14:23 | dtantsur | mordred: okay, I'll see what I can do | |
| 14:14:42 | dtroyer | mordred: coolio, will have a look | |
| 14:23:42 | dtroyer | mordred: looks nice, thank you! | |
| 14:27:57 | mordred | dtroyer: yay! | |
| 16:11:18 | openstackgerrit | Dmitry Tantsur proposed openstack/openstacksdk master: Add Resource._from_munch class method https://review.openstack.org/579908 | |
| 16:11:19 | dtantsur | mordred: ^^^ | |
| 16:11:28 | dtantsur | now let's see if I can get a good use of it | |
| 16:11:40 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Use shade create_image method in image proxy layer https://review.openstack.org/579168 | |
| 16:11:48 | mordred | dtantsur: neat. I could probably use that in that patch ^^ | |
| 16:12:52 | dtantsur | mordred: oh, curious, I'm doing it the other way around in my patch: shade level calling baremetal proxy | |
| 16:13:02 | dtantsur | is there The Right approach here? or just whatever fits? | |
| 16:20:02 | dtantsur | hmm, and should I wire this _from_munch into Proxy._get_resource? so that all proxies support it automagically? | |
| 16:20:14 | dtantsur | this can be dangerous, maybe only for ironic for now.. | |
| 16:22:47 | dtantsur | actually, from_munch can probably be implemented using the Body() fields | |
| 16:32:18 | openstackgerrit | Dmitry Tantsur proposed openstack/openstacksdk master: Add Resource._from_munch class method https://review.openstack.org/579908 | |
| 16:43:29 | openstackgerrit | Dmitry Tantsur proposed openstack/openstacksdk master: Handle Munch objects in proxies https://review.openstack.org/579908 | |
| 16:55:15 | openstackgerrit | Dmitry Tantsur proposed openstack/openstacksdk master: Handle Munch objects in proxies https://review.openstack.org/579908 | |
| 17:42:37 | mordred | dtantsur|afk: your way is the better way - we want to get to the shade layer calling the proxy layer rather than the other way around | |
| 17:43:37 | mordred | dtantsur|afk: but for create_image that's going to require some surgery, so I figured invert it for now, then iterate on getting it done right as a followup | |
| 20:28:53 | openstackgerrit | David Shrewsbury proposed openstack-infra/shade master: Fix for passing dict for get_* methods https://review.openstack.org/579974 | |
| 20:29:12 | Shrews | that probably could use a test | |
| 20:29:43 | mordred | yah. also - I'm pretty sure pep8 is going to complain about the whitespace | |
| 20:32:06 | openstackgerrit | David Shrewsbury proposed openstack-infra/shade master: Fix for passing dict for get_* methods https://review.openstack.org/579974 | |
| 20:32:18 | Shrews | pep8 passed for me | |
| 20:33:56 | mordred | Shrews: cool | |
| 20:35:31 | Shrews | mordred: you want a release note for that? | |
| 20:38:51 | wacuuu | yyy question: are there any plans for sdk in go? | |
| 20:42:17 | mordred | wacuuu: the gophercloud project is basically that | |
| 20:42:22 | mordred | Shrews: nah | |